IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
This doesn't make sense if leader broadcasts are used.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Now all references to ctdb->recovery_lock are encapsulated in the
cluster lock code.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
It is no longer just a recovery lock but is always held by the cluster
leader.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The introduction of the leader broadcast timeout provides an
alternative to the current leader validation. Using the leader
broadcast may not be as fast but it is more correct.
When the leader node is stopped or banned, the only way of triggering
an election is currently to fetch the leader's node map to check
whether the it is still active. This is because the leader will no
longer push the node map to other nodes. However, having all nodes
fetch the node map from an inactive leader may be unreliable.
Most of the other cases are also handled more reliably by the leader
broadcast timeout.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This no longer occurs at startup due to the leader broadcast timeout.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
If no leader broadcasts have been received from the leader for more
than 5s then trigger an election.
Apart from being sane behaviour, this avoids elected-before-connected
bugs at startup, where a node elects itself leader before it is
connected to other nodes.
When a node processes a leader broadcast timeout it sends an unknown
leader broadcast to all nodes. That causes cancellation of the leader
broadcast timeout across the cluster. This is particular important at
startup, since nodes may be started in a staggered fashion. Without
this cluster-wide cancellation, a node might notice the lack of
leader, win an election and complete a recovery before other nodes
notice the lack of leader. When the leader broadcast timeout finally
occurs on the other nodes then they'll put the cluster back into an
unnecessary recovery.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
These are triggered on 1 second timer, but are only sent if the node
is the current leader and there is no election underway.
If this node can not be the leader then ensure it releases the
recovery lock.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
An alternate election method will be added that doesn't use the
election timeout, so this provides a common way for recognising when
an election is in progress.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This makes the code self-documenting.
In ctdb_election_data() there is a slight behaviour change. An
inactive node will now try to lose an election. This case should not happen
because:
* An inactive node can't win an election round and then send a reply.
* Any inactive node should never start an election. There are
currently places where this happens and they will be fixed later.
There is an instance where this could be used in
validate_recovery_master() but this involves a more serious logic
change. Overhaul this function later.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
There are some remaining instances in this file but they will be
removed in subsequent commits.
Modernise debug macros as appropriate.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Recovery master is being renamed to leader. This follows clustering
best practice (e.g. RAFT).
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This is currently referenced in a number of inconsistent
ways, including:
* pnn
* rec->ctdb->pnn
* ctdb->pnn
* ctdb_get_pnn(ctdb)
* ctdb_get_pnn(rec->ctdb)
The first of these always requires some thought about the context - is
this the node PNN or some other PNN (e.g. argument to function)?
rec->pnn is now always used when referring to the recovery daemon's
PNN.
Doing this also reduces reliance on struct ctdb_context internals.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
ban_time argument is always ctdb->tunable.recovery_ban_period, so
build this in and make the calling code more readable.
ctdb_ban_node() already logs how long a node is banned for, so don't
repeatedly log this.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
All other arguments are available via rec, so simplify.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
pnn and nodemap are both available via the rec context, so simplify.
vnnmap is unused.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The pnn and nodemap arguments to force_election() and
send_election_request() are always effectively rec->pnn and
rec->nodemap, so simplify.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This is currently referenced in a number of inconsistent
ways, including:
* pnn
* rec->ctdb->pnn
* ctdb->pnn
* ctdb_get_pnn(ctdb)
* ctdb_get_pnn(rec->ctdb)
The first of these always requires some thought about the context - is
this the node PNN or some other PNN (e.g. argument to function)?
The intention is to always use rec->pnn when referring to the recovery
daemon's PNN.
Doing this also reduces reliance on struct ctdb_context internals.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Make the code self-documenting.
This preempts an upcoming change to terminology but doing it now saves
a lot of churn.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The recovery and takeover helpers can run for a while and generate
non-trivial logs, so have them reopen their logs to support log
rotation.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Mon Jan 17 04:36:30 UTC 2022 on sn-devel-184
Recovery and takeover helpers can run for a while and generate
non-trivial logs. They should support log reopening.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Pass on a SIGHUP to the recovery daemon, which will then reopen its
logs.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Remote nodes are already initialised as UNHEALTHY when the node list
is initialised at startup (ctdb_load_nodes_file() calls
convert_node_map_to_list()) and when disconnected (ctdb_node_dead()).
So, drop this code.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Thu Sep 9 02:38:34 UTC 2021 on sn-devel-184
If this node is not connected to a node then we shouldn't know
anything about it. The state will be pushed later by the recovery
master.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Now that there are separate disable/enable controls used by the ctdb
tool this control can ignore any flag updates for the current nodes.
These only come from the recovery master, which depends on being able
to fetch flags for all nodes.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
CTDB_SRVID_SET_NODE_FLAGS is no longer sent so drop monitor_handler()
and replace with srvid_not_implemented(). Mark the SRVID obsolete in
its comment.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The code that handles this message is
ctdb_recoverd.c:monitor_handler(). Although it appears to do
something potentially useful, it only logs the flags changes. All
changes made are to local structures - there are no actual
side-effects.
It used to trigger a takeover run when the DISABLED flag changed.
This was dropped back in commit
662f06de9f.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
When flags change, promote the message to NOTICE level and switch the
message to the style that is currently generated by
ctdb-recoverd.c:monitor_handler(). This will allow monitor_handler()
to go away in future.
Drop logging when flags do not change. The recovery master now logs
when it pushes flags for a node, so the lack of a corresponding
"changed flags" message here indicates that no update was required.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Don't trust the old flags from the recovery master.
Surrounding code will change in future comments, including the use of
old-style debug macros, so just make this change clear.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
DISABLED is UNHEALTHY | PERMANENTLY_DISABLED, which is not what is
intended here. Luckily, it doesn't do any harm because nodes are
marked unhealthy at startup anyway.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This will usually happen if flags on the node in question change, so
keeping the code simple and pushing to all nodes won't hurt. When all
nodes come up there might be differences in connected nodes, causing
such "fix ups". Receiving nodes will ignore no-op pushes.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The resulting code structure looks a little weird. However, there is
another condition that requires the flags to be pushed that will be
inserted before the continue statement in a subsequent commit..
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14784
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The socket is set close-on-exec but that doesn't help for processes
that do not exec(). This should be done for all child processes.
This has been seen in testing where "ctdb shutdown" waits for the
socket to close before succeeding. It appears that lingering
vacuuming processes have not closed the socket when becoming clients
so they cause "ctdb shutdown" to hang even though the main daemon
process has exited. The cause of the lingering vacuuming processes
has been previously examined but still isn't understood.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
1. PID of lock helper waiting for lock
2. Scope of lock: "record" or "db"
3. Path to database that lock helper is trying to lock
4. Whether the database uses mutexes: "mutex" or "fcntl"
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
If run with UID wrapper and UID_WRAPPER_ROOT=1 then securing the
socket will fail.
Test mode means that local daemons are in use, so securing the socket
is not important.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Volker Lendecke <vl@samba.org>
Variable res is only used once and ret is re-used many times. Drop
res, use ret, which doesn't need to be initialised. Modernise debug
macro.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Volker Lendecke <vl@samba.org>
update_flags() has already updated the recovery master's canonical
node map, based on the flags from each remote node, and pushed out
these flags to all nodes.
If i == j then the node map has already been updated from this remote
node's flags, so simply drop this case.
Although update_flags() has updated flags for all nodes, it did not
update each node map in remote_nodemaps[] to reflect this. This means
that remote_nodemaps[] may contain inconsistent flags for some nodes
so it should not be used to check consistency when i != j.
Further, a meaningful difference in flags can only really occur if
update_flags() failed. In that case this code is never reached.
These observations combine to imply that this whole loop should be
dropped.
This leaves potential sub-second inconsistencies due to out-of-band
healthy/unhealthy flag changes pushed via CTDB_SRVID_PUSH_NODE_FLAGS.
These updates could be dropped (takeover run asks each node for
available IPs rather than making centralised decisions based on node
flags) but for now they will be fixed in the next iteration of
main_loop().
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14513
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This has already been done in update_flags().
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14513
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This is no longer necessary because the capability new style database
pull is assumed to always be available.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Removes use of the old controls without cleaning up the code. Clean
up can be done later.
After this change the CTDB_CAP_FRAGMENTED_CONTROLS capability is no
longer checked. This capability can be removed along with the
controls.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This makes it consistent with the monitoring code. If the master has
changed then this means the master will always get the message.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Tue Aug 18 06:24:11 UTC 2020 on sn-devel-184
This also updates remote flags so the name is misleading.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
update_local_flags() will be changed to use these nodemaps.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The nodemap has already been fetched from the local node and is
actually passed to this function. Care must be taken to avoid
referencing the "remote" nodemap for the recovery master. It also
isn't useful to do so, since it would be the same nodemap.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The plan here is to use the nodemaps retrieved by get_remote_nodes()
in update_local_flags(). This will improve efficiency, since
get_remote_nodes() fetches flags from nodes in parallel. It also
means that get_remote_nodes() can be used exactly once early on in
main_loop() to retrieve remote nodemaps. Retrieving nodemaps multiple
times is unnecessary and racy - a single monitoring iteration should
not fetch flags multiple times and compare them.
This introduces a temporary behaviour change but it will be of no
consequence when the above changes are made.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This array is indexed by the same index as nodemap, not the PNN.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Also drop error handling in main_loop() that is replaced by this
change.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This will allow an error callback to be added.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Change 1st argument to a rec context, since this will be needed later.
Drop the nodemap argument and access it via rec->nodemap instead.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The memory is allocated off the memory context used by the current
iteration of main loop. It is freed when main loop completes the fix
doesn't require backporting to stable branches. However, it is sloppy
so it is worth fixing.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Don't log an error on failure - let the caller can do this. Apart
from this: fix up coding style and modernise the remaining error
message.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14466
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Mon Aug 3 22:21:04 UTC 2020 on sn-devel-184
The logic currently in ctdb_ctrl_modflags() will be optimised so that
it no longer matches the pattern for a control function. So, remove
this function and squash its functionality into the only caller.
Although there are some superficial changes, the behaviour is
unchanged.
Flattening the 2 functions produces some seriously weird logic for
setting the new flags, to the point where using ctdb_ctrl_modflags()
for this purpose now looks very strange. The weirdness will be
cleaned up in a subsequent commit.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This is clearer than using the MODFLAGS control directly.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This makes fields such as recmaster and nodemap easily available if
required.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
An unused argument needlessly extends the length of function calls. A
subsequent change will allow rec->nodemap to be used if necessary.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
The current code only ever swaps with slot 0. This will only ever
happen with slots 0 and 1, so probably never sorts.
Replace with qsort().
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
ctdbd currently only logs when a new hot key is added. If a key gets
hotter then nothing new is logged.
Log hot key updates when the number of migrations has doubled since
the last time that key was logged.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This message indicates that a hot key was added, so say that. After
all the hot key slots have been filled the id will always be 0, so
don't bother logging it.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
These should be unsigned but luck is currently on our side.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
There are 2 reasons for this. Sorting of hot keys is broken and will
be changed to an implementation that needs a named (i.e. not
anonymous) structure. Also, at least one non-protocol field will be
added to facilitate more useful logging.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Testing control: 4 bytes msec delay plus a blob, return the request after the
delay. This is an enhanced "ping" which can be used to test asynchronous
clients.
Doesn't have the full protocol implementation yet
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Martin Schwenke <martin@meltin.net>
The vacuuming integration tests set VacuumInterval to a very high
number to avoid vacuuming collisions. This is done after the cluster
is healthy, so Samba will have already been started and vacuuming will
already be scheduled *at the default interval* for databases attached
by Samba. This means that vacuuming controls used by vacuuming tests
can still collide with the scheduled vacuuming events.
Add some logic to reschedule a vacuuming event that has fired but
where VacuumInterval has increased since it was originally scheduled.
The increase in VacuumInterval is used as the time offset for
rescheduling the event.
Although this changes production behaviour for the convenience of
testing, the new behaviour is completely reasonable and obeys the
principle of least surprise.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Tue Apr 7 03:04:57 UTC 2020 on sn-devel-184
No behaviour change. This is final staging to make the next change
completely obvious.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
No behaviour change. This just makes future changes clearer by
avoiding reformatting (or introducing local variables).
Clean up error handling while touching a relevant line.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Inside the nested event loop in ctdb_ctrl_getnodemap(), various
asynchronous handlers may dereference rec->nodemap, which will be
NULL.
One example is lost_reclock_handler(), which causes rec->nodemap to be
unconditionally dereferenced in list_of_nodes() via this call chain:
list_of_nodes()
list_of_active_nodes()
set_recovery_mode()
force_election()
lost_reclock_handler()
Instead of attempting to trace all of the cases, just avoid leaving
rec->nodemap set to NULL. Attempting to use an old value is generally
harmless, especially since it will be the same as the new value in
most cases.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14324
Reported-by: Volker Lendecke <vl@samba.org>
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Martin Schwenke <martins@samba.org>
Autobuild-Date(master): Tue Mar 24 01:22:45 UTC 2020 on sn-devel-184
Neither the recovery daemon nor the recovery helper should attach
databases outside of the recovery process.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This builds a complete list of databases across the cluster so it can
be used to create databases on the nodes where they are missing.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This will be used to build a merged list of databases from all nodes,
allowing the recovery helper to create missing databases.
It would be possible to also include the db_name field in this
structure but that would cause a lot of churn. This field is used
locally in the recovery of each database so can continue to live in
the relevant state structure(s).
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This is currently only set by the recovery daemon when it attaches
missing databases, so there is no obvious behaviour change. However,
attaching missing databases can now be moved to the recovery helper as
long as it sets this flag.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
ctdb_ctrl_createdb() is only called by the recovery daemon, so this is
a safe, temporary change. This is temporary because
ctdb_ctrl_createdb(), create_missing_remote_databases() and
create_missing_local_databases() will all go away soon.
Note that this doesn't cause a change in behaviour. The main daemon
will still only defer attaches from non-recoverd processes during
recovery.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Commit 3cc230b5ee says:
Dont allow clients to connect to databases untile we are well past
and through the initial recovery phase
It is unclear what this commit was attempting to do. The commit
message implies that more attaches should be deferred but the code
change adds a conjunction that causes less attaches to be deferred.
In particular, no attaches will be deferred after startup is complete.
This seems wrong.
To implement what seems to be stated in the commit message an "or"
needs to be used so that non-recovery daemon attaches are deferred
either when in recovery or before startup is complete. Making this
change highlights that attaches need to be allowed during the
"startup" event because this is when smbd is started.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
If a node is marked for banning, confirm that it's not become inactive
during the recovery. If yes, then don't ban the node.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
It's possible to have a node stopped, but recovery master not yet
updated flags on the local ctdb daemon when recovery is started. So do
not trust the list of active nodes obtained from the local node. Query
the connected nodes to calculate the list of active nodes.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
New vnnmap is constructed using the information from all the connected
nodes. So there is no need to fetch the vnnmap from recovery master.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
If NODE_FLAGS_DISCONNECTED is set the node can be in half-connected state. With
this change we ensure to restart the transport for this case.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14295
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Martin Schwenke <martin@meltin.net>
There is no sane way of keeping stdin open when using the shell to
background ctdbd in local_daemons.sh. Instead, have ctdbd fork when
not interactive and when test mode is enabled. become_daemon() can't
be used for this: if it forks then it also closes stdin.
For the interactive case, become_daemon() wasn't doing anything
special, so do nothing instead.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
These don't need to depend on do_fork. Child logging should be set up
whenever the daemon is not interactive. The stdin handler should be
setup whenever test mode is enabled.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
No functional changes.
This is staging for a change that makes ctdbd fork when test mode is
enabled but interactive is not set.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This allows a test environment to simply close its end of a pipe to
cleanly shutdown ctdbd. Like in smbd, this is only done if stdin is a
pipe or a socket.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This avoids a crash if ctdb_shutdown_sequence() is called before
monitoring is initialised.
Switch to using TALLOC_FREE() while touching this function.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Testing against a commonly used cluster filesystem has shown no
performance impact, as expected.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This changes the behaviour for some failures from exiting to simply
attempting to schedule the next run.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
All the vacuum operations if required have an event loop to ensure
completion of pending operations. Once all the steps are complete,
there is no reason to process any more packets.
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
The only reason for recoverd attaching to databases was to migrate
records to the local node as part of vacuuming. Recovery daemon does
not take part in database vacuuming any more.
The actual database recovery is handled via the recovery_helper and
recovery daemon should not need to attach to the databases any more.
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
This is now implemented in the ctdb daemon using VACUMM_FETCH control.
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
This occurs rarely but can adversely impact performance, so it is
worth logging it more frequently.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This currently skips the last record.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14147
RN: Avoid potential data loss during recovery after vacuuming error
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
Currently some of this is supported by a periodic check in the
recovery daemon's main_loop(), which notices the flag change, sets
recovery mode active and freezes databases. If STOP_NODE returns
immediately then the associated recovery can complete and the node can
be continued before databases are actually frozen.
Instead, immediately do all of the things that make a node inactive.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14087
RN: Stop "ctdb stop" from completing before freezing databases
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Tue Aug 20 08:32:27 UTC 2019 on sn-devel-184
There's no reason to avoid immediately setting recovery mode to active
and initiating freeze of databases.
This effectively reverts the following commits:
d8f3b490bbb4357a79d9
The latter is now implemented using a control, resulting in looser
coupling.
See also the following commit:
f8141e91a6
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14087
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>